Landscape evaluation by using no-code machine-learning software

Case study of a suburban rural village in Hikone City, Shiga Prefecture, Japan

Abstract

I examined the construction of a simple landscape evaluation method that uses existing no-code machine-learning software. Specifically, I targeted rural areas near urban areas, extracted image data from footage captured with an action camera (GoPro), and then attempted image classification and object detection by using no-code machine-learning software. I was able to distinguish, with sufficient accuracy, the differences between traditional private houses in existing villages and houses created from subdivisions. If conditions such as the angle of the image differed, erroneous judgments could occur. However, by point visualization using object detection, I was able to infer which landscape elements influenced such misjudgments. In the conclusion to my report, I summarize the points of view that are essential to solving a series of issues, and I provide future prospects.